Microsoft Research Asia has been working on creating software called MoodScope that notes how a user uses his or her phone, and then uses that information to guess that user's mood. Initial testing of the device has shown it to be 66 percent accurate; when tailored to an individual user, the team reports that the accuracy rate jumped to 93 percent. The research team includes Nicholas Lane and Robert LiKamWa of Rice University, and Lin Zhong and Yunxin Liu from Microsoft Research Asia. They built a prototype and posted their test study results on Microsoft's website.
|The circumplex mood model. Credit: Robert LiKamWa et al.|
Most people realize that their smartphone has a lot of embedded technology in it that interacts with the world at large—GPS hardware, accelerometers, etc. all monitor activity and use that data to provide useful functions, such as automatically switching from landscape to portrait mode when a phone is rotated. In this new effort, the researchers sought to discover whether software that monitors phone activities could reveal the users' moods.
To find out, the team wrote code that monitored email, texting, app usage, phone calls, location information, and browsing history, then added algorithms to guess mood based on that data. Next, they enlisted the assistance of 32 volunteers to help them test the accuracy of their code. The volunteers were asked to use the system for two months while also completing mood assessments to provide data for comparison. With no training or tweaking, the software was found to provide answers of happy, tense, calm, upset, excited, stressed, or bored that matched the actual mood reported by the volunteers, on average 66 percent of the time. After optimizing the system for the individual habits of each of the volunteers, the rate increased to 93 percent.
The researchers suggest third party hooks could be added to the software to allow for automatically transmitting user moods to applications like Facebook. They also acknowledge that privacy concerns could arise if the software were to be delivered to the public, but suggest the benefits of such software would likely outweigh such concerns. They note that sites like Netflix or Spotify could use data from MoodScope to offer movies or other content based on specific users' moods.
The team presented their findings at MobiSys 2013 held in Taiwan last month.
More information: MoodScope Building a Mood Sensor from Smartphone Usage Patterns: research.microsoft.com/apps/pubs/default.aspx?id=194498
Research paper: www.ruf.rice.edu/~mobile/publications/likamwa2013mobisys2.pdf